Service providers are taking multiple routes to architecting and/or opening their 5G networks. Some are handing them off to hyperscalers, either for specific applications or more broadly. They face the decision of what data to keep under their own control on their own compute, where they control the latency and resiliency measures, and what they feel confident enough to keep with a cloud provider, she explains. And then there’s the issue of cost and how much expense it takes to deploy something.
“For service providers, they want to make whatever infrastructure they use … to be the cheapest way to deploy something. So if it’s cheaper to deploy with a hyperscaler, that’s what they’re going to do,” says Rimma Iontel, chief architect for the global telco team at Red Hat. “Some of them are basically looking at all of these considerations and pick two, maybe,or try to balance them against each other, to find the solution that suits them the best,” she adds.
As CSPs transition from legacy infrastructure to cloud-based infrastructure, though, even within their own data centers “that’s already a huge shift, because you have so many layers of abstraction,” she explains. “It makes it a lot harder to monitor each of those layers and more importantly, to correlate between those layers. When something fails, it’s sometimes really difficult to figure out quickly whether it’s a software failure, a hardware failure, a network failure. Because even in the network layer, you have abstractions between the underlay and the overlay, and you might not necessarily have the visibility to allow you to make that determination. So basically, your troubleshooting cycles could blow up tremendously.” Adding hyperscalers to the picture and the difference between visibility and control in bare-metal deployments (some control of the underlying hardware) versus infrastructure-as-a-service (typically a view of virtual machines rather than the underlying hardware layer), and visibility can vary greatly. “You have to learn, first of all, how to protect yourself against [that], to make it irrelevant, essentially, whether you can see it or not,” she adds. That’s the entire concept of such as-a-service platforms, after all – but, Iontel adds, “the applications have to be designed in such a way, and deployed in such a way, as to make that feasible.
“Some of the applications that we come across, even though they’ve been converted to virtualized environment, do not necessarily function as cloud native applications,” she adds. “So when service providers speak the vendors for the applications, that’s something they need to keep in mind and test for.”
And, she adds, a comprehensive solution, which has visibility across everything is “still a little bit of a wish, more than a reality. But our customers are learning, basically, how to break down the problem. So you’re not trying to drink the ocean, but you’re breaking it up into manageable problems – you have visibility into smaller chunks of your network, and you correlate between those.”
Iontel emphasizes that 1) it’s important to do the groundwork, and 2) CSPs don’t have to reinvent the wheel. Red Hat, she notes, works with NEM partners on 5G RAN and core enablement in a containerized, software-centric, orchestrated network platform. “If the platform is adopted by a service provider, deploying this application on the platform is not something you have to invent from scratch. It’s something that’s already been done in our labs, in our partner labs, or even in labs like TIP community labs,” she points out. “That’s an excellent place to test solutions before you actually put them in your network, to make sure that conceptually does work. So, do the groundwork before you are ready to put something in production and capitalize on the work that’s already being done by others as well.”
Building upon that, she urges telco players to share information about what they, and their partners have done. If they want to adopt open source technologies, telcos should be part of the “open” part of that. “Work out in the open, share information, learn from it and share those lessons with others,” she says. Red Hat makes a point to work with carriers as partners and talk publicly about its learnings, Iontel adds. “Our learnings, we publish them. We talk about them in industry forums. We bring them to standardization bodies, so others can benefit from the work that’s been done.” Amid a shifting telco network environment, she urges network operators and their various partners to do the same.
